PyOVERCAST 1.0.13__tar.gz → 1.0.15__tar.gz

{pyovercast-1.0.13 → pyovercast-1.0.15}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: PyOVERCAST
-Version: 1.0.13
+Version: 1.0.15
 Summary: A Python package for mining key transcription factors from transcriptome data.
 Author-email: Tinghua Huang <thua45@126.com>
 License-Expression: MIT

{pyovercast-1.0.13 → pyovercast-1.0.15}/PyOVERCAST/predict.py

@@ -14,6 +14,7 @@ plt.rcParams['axes.unicode_minus'] = False  # 解决负号显示问题
 import multiprocessing as mp
 import seaborn as sns
 from scipy import stats
+from scipy.stats import false_discovery_control
 import random
 
 overcast_data_dir = './OVERCAST_data'
@@ -104,7 +105,7 @@ def Z_to_frame(Z):
 	return df
 
 def compute_dtci_core(df, x_col='x', y_col='y', z_col='z',
-					  c=1.0, alpha=1.0, beta_w=0.5, gamma=1.0):
+					  c=1.0, alpha=1.0, beta_w=0.5, gamma=1.0, m=2.0):
 	"""
 	核心DTCI计算 —— 使用 statsmodels.OLS
 
@@ -127,6 +128,7 @@ def compute_dtci_core(df, x_col='x', y_col='y', z_col='z',
 	# 使用 statsmodels.OLS（自动处理常数项、标准误和p值）
 	X_design = sm.add_constant(pd.DataFrame({'x': x, 'y': y, 'xy': xy}))
 	model = sm.OLS(z, X_design).fit()
+	rmse = np.sqrt(model.mse_resid)
 
 	# 提取系数 [const, x, y, xy]
 	beta_hat = model.params.values
@@ -143,7 +145,16 @@ def compute_dtci_core(df, x_col='x', y_col='y', z_col='z',
 	z_std = np.std(z, ddof=1)
 
 	# 子指标
-	DI = np.abs(beta3) / (np.abs(beta3) + c * z_std)
+	# DI = np.abs(beta3) / (np.abs(beta3) + c * z_std)
+	# RI: 翘起深度指数 —— 修正版 v4.0
+	# 信噪比 = 超额高度 / 纯噪声（残差标准差）
+	# SNR = h_diff / rmse if rmse > 1e-10 else 0.0
+	SNR = np.abs(beta3) / rmse if rmse > 1e-10 else 0.0
+	# Sigmoid 映射：SNR=m 时 RI=0.5，SNR 越大 RI 越接近1
+	if SNR > 0:
+		DI = 1.0 / (1.0 + np.exp(m - SNR))
+	else:
+		DI = 0.0
 	SI = min(1.0, -np.log10(max(p_values[3], 1e-10)) / 10.0)
 	QI = max(0.0, R2_adj)
 
@@ -706,6 +717,8 @@ def olcr_bootstrap(set_names=None, list_file=None, win=30, thread_n=16, bs_n=200
 		results.append(rline)
 	col_names = ['TF_motif', 'Beta3', 'Beta3_Pvalue', 'DTCI', 'DTCI_Pvalue', 'Rank_Score', 'Rank / N']
 	df = pd.DataFrame(results, columns=col_names)
+	fdr = false_discovery_control(df['DTCI_Pvalue'], method='bh')
+	df.insert(loc=5, column='DTCI_FDR', value=fdr)
 
 	return df
 
@@ -899,11 +912,13 @@ def plot_contour_graph(rrho_df, out_file=None):
 	ax.set_title('Contour Graph of RRLO Matrix')
 	# ax.set_aspect('equal')
 	plt.tight_layout()
-	#plt.savefig('contour_graph_invF.png', dpi=150)
-	plt.show()
+	if out_file != None:
+		plt.savefig(out_file, dpi=150)
+	else:
+		plt.show()
 
 
-def plot_contour_plot(df1, df2, win):
+def plot_contour_plot(df1, df2, win, out_file=None):
 	gene_col1 = df1.columns[0]
 	score_col1 = df1.columns[1]
 	df1 = df1.sort_values(score_col1, ascending=False)
@@ -930,9 +945,9 @@ def plot_contour_plot(df1, df2, win):
 
 	#beta_xy, pval_xy = td_coor(rrho_df.to_numpy())
 	# df = Z_to_frame(rrho_df.to_numpy())
-	plot_contour_graph(rrho_df)
+	plot_contour_graph(rrho_df, out_file=out_file)
 
-def plot_contour(set_names=['human_hocomoco_CLink_wtcoor_1w_0.8'], list_file='input_deg-list.txt', tf='NFKB1_HUMAN.H11MO.1.B', win=30, data_dir=None):
+def plot_contour(set_names=['human_hocomoco_CLink_wtcoor_1w_0.8'], list_file='input_deg-list.txt', tf='NFKB1_HUMAN.H11MO.1.B', win=30, data_dir=None, out_file=None):
 	if data_dir == None:
 		data_dir = overcast_data_dir
 	if not (os.path.exists(data_dir) and os.path.isdir(data_dir)):
@@ -958,10 +973,10 @@ def plot_contour(set_names=['human_hocomoco_CLink_wtcoor_1w_0.8'], list_file='in
 	epsilon = 1e-15   # 常用值
 	for i in range(tf_n):
 		if tfbs_array[i][0] == tf:
-			plot_contour_plot(deg_list, tfbs_array[i][1], win)
+			plot_contour_plot(deg_list, tfbs_array[i][1], win, out_file=out_file)
 			break
 
-def plot_contour_permutation(set_names=['human_hocomoco_CLink_wtcoor_1w_0.8'], list_file='input_deg-list.txt', tf='NFKB1_HUMAN.H11MO.1.B', win=30, data_dir=None):
+def plot_contour_permutation(set_names=['human_hocomoco_CLink_wtcoor_1w_0.8'], list_file='input_deg-list.txt', tf='NFKB1_HUMAN.H11MO.1.B', win=30, data_dir=None, out_file=None):
 	if data_dir == None:
 		data_dir = overcast_data_dir
 	if not (os.path.exists(data_dir) and os.path.isdir(data_dir)):
@@ -988,7 +1003,7 @@ def plot_contour_permutation(set_names=['human_hocomoco_CLink_wtcoor_1w_0.8'], l
 	for i in range(tf_n):
 		if tfbs_array[i][0] == tf:
 			gene_set_pt, gene_list_pt = permutation(tfbs_array[i][1], deg_list)
-			plot_contour_plot(gene_list_pt, gene_set_pt, win)
+			plot_contour_plot(gene_list_pt, gene_set_pt, win, out_file=out_file)
 			# plot_contour_plot(deg_list, tfbs_array[i][1], win)
 			break
 

{pyovercast-1.0.13 → pyovercast-1.0.15}/PyOVERCAST.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: PyOVERCAST
-Version: 1.0.13
+Version: 1.0.15
 Summary: A Python package for mining key transcription factors from transcriptome data.
 Author-email: Tinghua Huang <thua45@126.com>
 License-Expression: MIT

{pyovercast-1.0.13 → pyovercast-1.0.15}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "PyOVERCAST"
-version = "1.0.13"
+version = "1.0.15"
 license = "MIT"  # SPDX expression
 description = "A Python package for mining key transcription factors from transcriptome data."
 readme = "README.md"